This September, our PhD students made waves at two major European conferences, bringing home a Best Paper Award and advancing the conversation on how machine learning can tackle real-world challenges.
A busy week for Machine Learning research
Silke Kaiser and Chiara Fusar Bassini from the AI and Climate Technology Policy Group represented Hertie School at the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) – Europe's flagship machine learning and data mining conference – held this year in Porto, Portugal. The conference buzzed with peer-reviewed research presentations, workshops, tutorials, and the kind of networking that (as Silke notes) “kept the conversations going long after the sessions ended.”
But the effort paid off. At the SoGood workshop, Silke's presentation on "Spatio-Temporal Graph Neural Network for Urban Spaces: Interpolating Citywide Traffic Volume" received the Best Paper Award. Not content with just one conference, Chiara also presented at the DACH+ Energy Informatics Conference in Aachen, Germany – two major ML conferences in one week!
What they're working on
Chiara's research: unlocking energy flexibility
Chiara uses machine learning to identify bottlenecks in the energy transition. Her work presented at ECML PKDD focused on revealing the flexibility (or lack thereof) of gas-fired generation units through deep clustering of electricity generation time series. "The application of data-driven methods to climate-related topics can help us understand them, so we can best address them," she explains.
Silke's research: decoding urban traffic
Silke investigates "how machine learning can help us understand spatio-temporal data — in plain words: patterns that happen across space and time." Her award-winning work focuses specifically on urban transport. She explains: "At ECML-PKDD I presented a paper where I developed a Graph Neural Network to predict city-wide traffic volumes. Graph Neural Networks combine the strengths of machine learning with the structure of networks. And since a street network is basically just a giant graph, it's the perfect playground."
Beyond the data: why they do what they do
Chiara discovered machine learning during her bachelor in Economics, Management and Computer Science: "I had always been deeply fascinated by discrete mathematics, and wanted to learn more on its applications - that's how I ended up learning about quantitative methods." Her motivation is clear: "I want to use machine learning and data science to tackle crucial societal challenges, such as the energy transition. With the Earth getting warmer year by year, contributing to design renewable-driven and efficient energy markets feels like a pretty relevant cause."
For Silke, "machine learning feels like logic, dressed up in statistics and supercharged with computing." Her drive is personal: "What motivates me is that I can apply these tools to purpose-driven topics, like transport. Cities are where we live, but also where we get annoyed by traffic almost every day. If I can help design models that improve transport experiences across modes, that's a pretty good motivator."
The reality of ML research
When asked about memorable ML moments, both researchers kept it real. Chiara offered "a pretty awesome picture of Porto in the early morning light" – practical advice that the city is definitely worth a visit.
Silke: "Nothing spectacular comes to mind right now either, which I know sounds like feeding the stereotype that “science is no fun.” In reality, it’s more about the small, everyday moments: like when your ML model confidently predicts something that makes absolutely no sense, or when you finally track down a stubborn bug only to realize it was your own typo all along."
Can ML make the world better?
Both researchers share a grounded perspective on machine learning's potential.
Chiara: “Machine Learning is not inherently "good" or "bad". It is a tool, so its final impact depends on how and why it is used. Nonetheless, I believe that policymaking can be significantly improved by using insights from machine learning and data science. That's where I see the role of research, making a tangible positive impact”.
Silke: “I completely agree with Chiara. ML on its own won’t save the world, but it can help us understand it better; whether that’s climate systems, transport networks, or energy grids. The “better place” part comes when humans decide to use those insights responsibly”.
Read more about Chiara’s paper “Seasonal and weather influences in the outages of German thermal generators” here and “Revealing the empirical flexibility of gas units through deep clustering” here.
